Connect HubSpot to your AI agent and ask about pipeline

1. Connect the HubSpot account you want to analyze

Open Connected Apps in Combined, choose Add source and search for HubSpot. Select Connect myself if you manage that HubSpot account. The authorization flow collects the credentials required by the current connector. If someone else owns the account, use the authorization request option so they can approve it themselves.

For a workspace you control, start with Private App authentication. In HubSpot’s settings, open Integrations → Private Apps and use a dedicated app for this connection. For the deals dataset, the connector documentation lists crm.objects.deals.read and crm.schemas.deals.read. Choose the scopes needed for the datasets you intend to sync.

Enter the Private App access token in Combined’s secure authorization form. Keep it out of the analysis prompt: the data connector uses the HubSpot credential, while the AI agent uses its own Combined access. If you choose OAuth credentials instead, follow the connector’s fields for client ID, client secret and refresh token.

2. Make the deals available to your agent

After authorization, select the deals dataset and the fields needed for your analysis. Keep its identifiers and update fields, along with deal name, amount, pipeline and stage. Run the first sync, then check the source’s last successful sync time and the available records in Explore.

For an AI-led setup, Combined’s Set up page supplies account-specific instructions for your agent. Follow those instructions through source configuration, access and the first query. To use an existing Claude Code workflow, follow the Combined MCP connection guide and grant access to the HubSpot data it needs.

Start with one business question and one dataset. You can add companies, contacts or other supported records when the question needs them. Combined’s cardless start includes 5 million MAR, so you can begin without choosing a paid plan first.

3. Ask a question with a checkable answer

Let’s use four sample deals, all in USD and in one sales pipeline, to make the question concrete.

Once your own data is ready, give your connected agent this instruction:

Use my connected HubSpot source in Combined.
Discover the deals dataset and its field names first.

What is our current open pipeline, grouped by stage?
Exclude stages marked closed-won or closed-lost.
Report the deal count and unweighted amount for each open stage,
then the overall total. Keep different currencies separate.
If stage IDs cannot be mapped, ask me for the stage mapping.
Flag missing amounts instead of silently treating them as zero.

Use a read-only aggregate query, not a sample of deals.
Include the last successful sync time, query receipt ID
and any truncation or coverage limits.

The stage mapping matters when your pipeline uses custom stages. Give the agent the relevant mapping if it is not available in the granted data. Let it discover logical relation and field names rather than copying a table name from someone else’s workspace.

4. Check the total, then ask the next question

These example deals add up to $20,000 in open pipeline across two deals. Qualified To Buy accounts for $12,000 and Presentation Scheduled for $8,000. The $5,000 won deal and $3,000 lost deal are excluded. This is an unweighted pipeline total, not booked revenue or a probability-weighted forecast.

Check your real result against the equivalent HubSpot view using the same pipeline, stage and currency filters. The answer describes the records as of the last successful sync. If someone has just edited a deal in HubSpot, sync that change before expecting it in the answer. See how to check agent data freshness for the details.

Now turn the total into a useful follow-up: “Which open deals are due to close this month?” Add the close date field and a precise date range. For “Which deals have stalled?”, first define stalled—such as no stage movement for 30 days—and make sure the required history exists. A current deal snapshot alone does not establish a history of stage changes.

Stop exporting spreadsheets every time someone has a pipeline question. Connect HubSpot once, keep the data synced, and ask your agent. Start with deals today; add the next dataset when your next question needs it.

Sources and further reading

Explore the documentation behind this guide. Product details checked on September 9, 2026.

• Combined connector authorization and sync
• Combined account onboarding
• HubSpot connector fields and required scopes
• HubSpot private app setup